# phrases.txt — the mini-corpus for the embeddings explorer.
# One phrase per line; lines starting with # are ignored.
# The corpus deliberately contains paraphrase pairs (same meaning, different
# words) and keyword-collision pairs (shared words, different meaning).

# — account access (paraphrase cluster) —
reset my password
credential recovery for a locked account
I forgot my login and can't get into my account
help me regain access to my user profile

# — phishing (paraphrase cluster) —
suspicious email asking me to click a link
possible phishing message received this morning
someone sent me a fake invoice with a strange attachment

# — malware / incident (paraphrase cluster) —
my laptop is infected with malicious software
ransomware encrypted the files on the workstation
antivirus flagged a trojan on the endpoint

# — network operations —
configure the firewall to block outbound traffic
open a port on the corporate VPN gateway
the wifi in the conference room keeps dropping

# — keyword-collision decoys (share words with clusters above, differ in meaning) —
the password to the wifi is posted in the break room
click the link below to view the lunch menu
recover deleted photos from my phone
the invoice for the new firewall hardware arrived

# — unrelated fillers —
book a meeting room for Friday afternoon
what time does the cafeteria close today
order more coffee for the third floor kitchen
